2 Copyright (C) 1996, 1997, 1998, 1999, 2000, 2001, 2002, 2003,
3 2004, 2005, 2006, 2007, 2008 Free Software Foundation, Inc.
5 This file is part of GNU Wget.
7 GNU Wget is free software; you can redistribute it and/or modify
8 it under the terms of the GNU General Public License as published by
9 the Free Software Foundation; either version 3 of the License, or (at
10 your option) any later version.
12 GNU Wget is distributed in the hope that it will be useful,
13 but WITHOUT ANY WARRANTY; without even the implied warranty of
14 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
15 GNU General Public License for more details.
17 You should have received a copy of the GNU General Public License
18 along with Wget. If not, see <http://www.gnu.org/licenses/>.
20 Additional permission under GNU GPL version 3 section 7
22 If you modify this program, or any covered work, by linking or
23 combining it with the OpenSSL project's OpenSSL library (or a
24 modified version of that library), containing parts covered by the
25 terms of the OpenSSL or SSLeay licenses, the Free Software Foundation
26 grants you additional permission to convey the resulting work.
27 Corresponding Source for a non-source form of such a combination
28 shall include the source code for the parts of OpenSSL used as well
29 as that of the covered work. */
44 #include "host.h" /* for is_valid_ipv6_address */
48 #endif /* def __VMS */
55 scm_disabled = 1, /* for https when OpenSSL fails to init. */
56 scm_has_params = 2, /* whether scheme has ;params */
57 scm_has_query = 4, /* whether scheme has ?query */
58 scm_has_fragment = 8 /* whether scheme has #fragment */
63 /* Short name of the scheme, such as "http" or "ftp". */
65 /* Leading string that identifies the scheme, such as "https://". */
66 const char *leading_string;
67 /* Default port of the scheme when none is specified. */
73 /* Supported schemes: */
74 static struct scheme_data supported_schemes[] =
76 { "http", "http://", DEFAULT_HTTP_PORT, scm_has_query|scm_has_fragment },
78 { "https", "https://", DEFAULT_HTTPS_PORT, scm_has_query|scm_has_fragment },
80 { "ftp", "ftp://", DEFAULT_FTP_PORT, scm_has_params|scm_has_fragment },
86 /* Forward declarations: */
88 static bool path_simplify (enum url_scheme, char *);
90 /* Support for escaping and unescaping of URL strings. */
92 /* Table of "reserved" and "unsafe" characters. Those terms are
93 rfc1738-speak, as such largely obsoleted by rfc2396 and later
94 specs, but the general idea remains.
96 A reserved character is the one that you can't decode without
97 changing the meaning of the URL. For example, you can't decode
98 "/foo/%2f/bar" into "/foo///bar" because the number and contents of
99 path components is different. Non-reserved characters can be
100 changed, so "/foo/%78/bar" is safe to change to "/foo/x/bar". The
101 unsafe characters are loosely based on rfc1738, plus "$" and ",",
102 as recommended by rfc2396, and minus "~", which is very frequently
103 used (and sometimes unrecognized as %7E by broken servers).
105 An unsafe character is the one that should be encoded when URLs are
106 placed in foreign environments. E.g. space and newline are unsafe
107 in HTTP contexts because HTTP uses them as separator and line
108 terminator, so they must be encoded to %20 and %0A respectively.
109 "*" is unsafe in shell context, etc.
111 We determine whether a character is unsafe through static table
112 lookup. This code assumes ASCII character set and 8-bit chars. */
115 /* rfc1738 reserved chars + "$" and ",". */
118 /* rfc1738 unsafe chars, plus non-printables. */
122 #define urlchr_test(c, mask) (urlchr_table[(unsigned char)(c)] & (mask))
123 #define URL_RESERVED_CHAR(c) urlchr_test(c, urlchr_reserved)
124 #define URL_UNSAFE_CHAR(c) urlchr_test(c, urlchr_unsafe)
126 /* Shorthands for the table: */
127 #define R urlchr_reserved
128 #define U urlchr_unsafe
131 static const unsigned char urlchr_table[256] =
133 U, U, U, U, U, U, U, U, /* NUL SOH STX ETX EOT ENQ ACK BEL */
134 U, U, U, U, U, U, U, U, /* BS HT LF VT FF CR SO SI */
135 U, U, U, U, U, U, U, U, /* DLE DC1 DC2 DC3 DC4 NAK SYN ETB */
136 U, U, U, U, U, U, U, U, /* CAN EM SUB ESC FS GS RS US */
137 U, 0, U, RU, R, U, R, 0, /* SP ! " # $ % & ' */
138 0, 0, 0, R, R, 0, 0, R, /* ( ) * + , - . / */
139 0, 0, 0, 0, 0, 0, 0, 0, /* 0 1 2 3 4 5 6 7 */
140 0, 0, RU, R, U, R, U, R, /* 8 9 : ; < = > ? */
141 RU, 0, 0, 0, 0, 0, 0, 0, /* @ A B C D E F G */
142 0, 0, 0, 0, 0, 0, 0, 0, /* H I J K L M N O */
143 0, 0, 0, 0, 0, 0, 0, 0, /* P Q R S T U V W */
144 0, 0, 0, RU, U, RU, U, 0, /* X Y Z [ \ ] ^ _ */
145 U, 0, 0, 0, 0, 0, 0, 0, /* ` a b c d e f g */
146 0, 0, 0, 0, 0, 0, 0, 0, /* h i j k l m n o */
147 0, 0, 0, 0, 0, 0, 0, 0, /* p q r s t u v w */
148 0, 0, 0, U, U, U, 0, U, /* x y z { | } ~ DEL */
150 U, U, U, U, U, U, U, U, U, U, U, U, U, U, U, U,
151 U, U, U, U, U, U, U, U, U, U, U, U, U, U, U, U,
152 U, U, U, U, U, U, U, U, U, U, U, U, U, U, U, U,
153 U, U, U, U, U, U, U, U, U, U, U, U, U, U, U, U,
155 U, U, U, U, U, U, U, U, U, U, U, U, U, U, U, U,
156 U, U, U, U, U, U, U, U, U, U, U, U, U, U, U, U,
157 U, U, U, U, U, U, U, U, U, U, U, U, U, U, U, U,
158 U, U, U, U, U, U, U, U, U, U, U, U, U, U, U, U,
164 /* URL-unescape the string S.
166 This is done by transforming the sequences "%HH" to the character
167 represented by the hexadecimal digits HH. If % is not followed by
168 two hexadecimal digits, it is inserted literally.
170 The transformation is done in place. If you need the original
171 string intact, make a copy before calling this function. */
174 url_unescape (char *s)
176 char *t = s; /* t - tortoise */
177 char *h = s; /* h - hare */
189 /* Do nothing if '%' is not followed by two hex digits. */
190 if (!h[1] || !h[2] || !(c_isxdigit (h[1]) && c_isxdigit (h[2])))
192 c = X2DIGITS_TO_NUM (h[1], h[2]);
193 /* Don't unescape %00 because there is no way to insert it
194 into a C string without effectively truncating it. */
204 /* The core of url_escape_* functions. Escapes the characters that
205 match the provided mask in urlchr_table.
207 If ALLOW_PASSTHROUGH is true, a string with no unsafe chars will be
208 returned unchanged. If ALLOW_PASSTHROUGH is false, a freshly
209 allocated string will be returned in all cases. */
212 url_escape_1 (const char *s, unsigned char mask, bool allow_passthrough)
219 for (p1 = s; *p1; p1++)
220 if (urlchr_test (*p1, mask))
221 addition += 2; /* Two more characters (hex digits) */
224 return allow_passthrough ? (char *)s : xstrdup (s);
226 newlen = (p1 - s) + addition;
227 newstr = xmalloc (newlen + 1);
233 /* Quote the characters that match the test mask. */
234 if (urlchr_test (*p1, mask))
236 unsigned char c = *p1++;
238 *p2++ = XNUM_TO_DIGIT (c >> 4);
239 *p2++ = XNUM_TO_DIGIT (c & 0xf);
244 assert (p2 - newstr == newlen);
250 /* URL-escape the unsafe characters (see urlchr_table) in a given
251 string, returning a freshly allocated string. */
254 url_escape (const char *s)
256 return url_escape_1 (s, urlchr_unsafe, false);
259 /* URL-escape the unsafe and reserved characters (see urlchr_table) in
260 a given string, returning a freshly allocated string. */
263 url_escape_unsafe_and_reserved (const char *s)
265 return url_escape_1 (s, urlchr_unsafe|urlchr_reserved, false);
268 /* URL-escape the unsafe characters (see urlchr_table) in a given
269 string. If no characters are unsafe, S is returned. */
272 url_escape_allow_passthrough (const char *s)
274 return url_escape_1 (s, urlchr_unsafe, true);
277 /* Decide whether the char at position P needs to be encoded. (It is
278 not enough to pass a single char *P because the function may need
279 to inspect the surrounding context.)
281 Return true if the char should be escaped as %XX, false otherwise. */
284 char_needs_escaping (const char *p)
288 if (c_isxdigit (*(p + 1)) && c_isxdigit (*(p + 2)))
291 /* Garbled %.. sequence: encode `%'. */
294 else if (URL_UNSAFE_CHAR (*p) && !URL_RESERVED_CHAR (*p))
300 /* Translate a %-escaped (but possibly non-conformant) input string S
301 into a %-escaped (and conformant) output string. If no characters
302 are encoded or decoded, return the same string S; otherwise, return
303 a freshly allocated string with the new contents.
305 After a URL has been run through this function, the protocols that
306 use `%' as the quote character can use the resulting string as-is,
307 while those that don't can use url_unescape to get to the intended
308 data. This function is stable: once the input is transformed,
309 further transformations of the result yield the same output.
311 Let's discuss why this function is needed.
313 Imagine Wget is asked to retrieve `http://abc.xyz/abc def'. Since
314 a raw space character would mess up the HTTP request, it needs to
315 be quoted, like this:
317 GET /abc%20def HTTP/1.0
319 It would appear that the unsafe chars need to be quoted, for
320 example with url_escape. But what if we're requested to download
321 `abc%20def'? url_escape transforms "%" to "%25", which would leave
322 us with `abc%2520def'. This is incorrect -- since %-escapes are
323 part of URL syntax, "%20" is the correct way to denote a literal
324 space on the Wget command line. This leads to the conclusion that
325 in that case Wget should not call url_escape, but leave the `%20'
326 as is. This is clearly contradictory, but it only gets worse.
328 What if the requested URI is `abc%20 def'? If we call url_escape,
329 we end up with `/abc%2520%20def', which is almost certainly not
330 intended. If we don't call url_escape, we are left with the
331 embedded space and cannot complete the request. What the user
332 meant was for Wget to request `/abc%20%20def', and this is where
333 reencode_escapes kicks in.
335 Wget used to solve this by first decoding %-quotes, and then
336 encoding all the "unsafe" characters found in the resulting string.
337 This was wrong because it didn't preserve certain URL special
338 (reserved) characters. For instance, URI containing "a%2B+b" (0x2b
339 == '+') would get translated to "a%2B%2Bb" or "a++b" depending on
340 whether we considered `+' reserved (it is). One of these results
341 is inevitable because by the second step we would lose information
342 on whether the `+' was originally encoded or not. Both results
343 were wrong because in CGI parameters + means space, while %2B means
344 literal plus. reencode_escapes correctly translates the above to
345 "a%2B+b", i.e. returns the original string.
347 This function uses a modified version of the algorithm originally
348 proposed by Anon Sricharoenchai:
350 * Encode all "unsafe" characters, except those that are also
351 "reserved", to %XX. See urlchr_table for which characters are
354 * Encode the "%" characters not followed by two hex digits to
357 * Pass through all other characters and %XX escapes as-is. (Up to
358 Wget 1.10 this decoded %XX escapes corresponding to "safe"
359 characters, but that was obtrusive and broke some servers.)
363 "http://abc.xyz/%20%3F%%36%31%25aa% a?a=%61+a%2Ba&b=b%26c%3Dc"
365 "http://abc.xyz/%20%3F%25%36%31%25aa%25%20a?a=%61+a%2Ba&b=b%26c%3Dc"
369 "foo bar" -> "foo%20bar"
370 "foo%20bar" -> "foo%20bar"
371 "foo %20bar" -> "foo%20%20bar"
372 "foo%%20bar" -> "foo%25%20bar" (0x25 == '%')
373 "foo%25%20bar" -> "foo%25%20bar"
374 "foo%2%20bar" -> "foo%252%20bar"
375 "foo+bar" -> "foo+bar" (plus is reserved!)
376 "foo%2b+bar" -> "foo%2b+bar" */
379 reencode_escapes (const char *s)
385 int encode_count = 0;
387 /* First pass: inspect the string to see if there's anything to do,
388 and to calculate the new length. */
389 for (p1 = s; *p1; p1++)
390 if (char_needs_escaping (p1))
394 /* The string is good as it is. */
395 return (char *) s; /* C const model sucks. */
398 /* Each encoding adds two characters (hex digits). */
399 newlen = oldlen + 2 * encode_count;
400 newstr = xmalloc (newlen + 1);
402 /* Second pass: copy the string to the destination address, encoding
403 chars when needed. */
408 if (char_needs_escaping (p1))
410 unsigned char c = *p1++;
412 *p2++ = XNUM_TO_DIGIT (c >> 4);
413 *p2++ = XNUM_TO_DIGIT (c & 0xf);
419 assert (p2 - newstr == newlen);
423 /* Returns the scheme type if the scheme is supported, or
424 SCHEME_INVALID if not. */
427 url_scheme (const char *url)
431 for (i = 0; supported_schemes[i].leading_string; i++)
432 if (0 == strncasecmp (url, supported_schemes[i].leading_string,
433 strlen (supported_schemes[i].leading_string)))
435 if (!(supported_schemes[i].flags & scm_disabled))
436 return (enum url_scheme) i;
438 return SCHEME_INVALID;
441 return SCHEME_INVALID;
444 #define SCHEME_CHAR(ch) (c_isalnum (ch) || (ch) == '-' || (ch) == '+')
446 /* Return 1 if the URL begins with any "scheme", 0 otherwise. As
447 currently implemented, it returns true if URL begins with
451 url_has_scheme (const char *url)
455 /* The first char must be a scheme char. */
456 if (!*p || !SCHEME_CHAR (*p))
459 /* Followed by 0 or more scheme chars. */
460 while (*p && SCHEME_CHAR (*p))
462 /* Terminated by ':'. */
467 scheme_default_port (enum url_scheme scheme)
469 return supported_schemes[scheme].default_port;
473 scheme_disable (enum url_scheme scheme)
475 supported_schemes[scheme].flags |= scm_disabled;
478 /* Skip the username and password, if present in the URL. The
479 function should *not* be called with the complete URL, but with the
480 portion after the scheme.
482 If no username and password are found, return URL. */
485 url_skip_credentials (const char *url)
487 /* Look for '@' that comes before terminators, such as '/', '?',
489 const char *p = (const char *)strpbrk (url, "@/?#;");
495 /* Parse credentials contained in [BEG, END). The region is expected
496 to have come from a URL and is unescaped. */
499 parse_credentials (const char *beg, const char *end, char **user, char **passwd)
505 return false; /* empty user name */
507 colon = memchr (beg, ':', end - beg);
509 return false; /* again empty user name */
513 *passwd = strdupdelim (colon + 1, end);
515 url_unescape (*passwd);
522 *user = strdupdelim (beg, userend);
523 url_unescape (*user);
527 /* Used by main.c: detect URLs written using the "shorthand" URL forms
528 originally popularized by Netscape and NcFTP. HTTP shorthands look
531 www.foo.com[:port]/dir/file -> http://www.foo.com[:port]/dir/file
532 www.foo.com[:port] -> http://www.foo.com[:port]
534 FTP shorthands look like this:
536 foo.bar.com:dir/file -> ftp://foo.bar.com/dir/file
537 foo.bar.com:/absdir/file -> ftp://foo.bar.com//absdir/file
539 If the URL needs not or cannot be rewritten, return NULL. */
542 rewrite_shorthand_url (const char *url)
547 if (url_scheme (url) != SCHEME_INVALID)
550 /* Look for a ':' or '/'. The former signifies NcFTP syntax, the
552 p = strpbrk (url, ":/");
556 /* If we're looking at "://", it means the URL uses a scheme we
557 don't support, which may include "https" when compiled without
558 SSL support. Don't bogusly rewrite such URLs. */
559 if (p && p[0] == ':' && p[1] == '/' && p[2] == '/')
564 /* Colon indicates ftp, as in foo.bar.com:path. Check for
565 special case of http port number ("localhost:10000"). */
566 int digits = strspn (p + 1, "0123456789");
567 if (digits && (p[1 + digits] == '/' || p[1 + digits] == '\0'))
570 /* Turn "foo.bar.com:path" to "ftp://foo.bar.com/path". */
571 ret = aprintf ("ftp://%s", url);
572 ret[6 + (p - url)] = '/';
577 /* Just prepend "http://" to URL. */
578 ret = aprintf ("http://%s", url);
583 static void split_path (const char *, char **, char **);
585 /* Like strpbrk, with the exception that it returns the pointer to the
586 terminating zero (end-of-string aka "eos") if no matching character
590 strpbrk_or_eos (const char *s, const char *accept)
592 char *p = strpbrk (s, accept);
594 p = strchr (s, '\0');
598 /* Turn STR into lowercase; return true if a character was actually
602 lowercase_str (char *str)
604 bool changed = false;
606 if (c_isupper (*str))
609 *str = c_tolower (*str);
615 init_seps (enum url_scheme scheme)
617 static char seps[8] = ":/";
619 int flags = supported_schemes[scheme].flags;
621 if (flags & scm_has_params)
623 if (flags & scm_has_query)
625 if (flags & scm_has_fragment)
631 static const char *parse_errors[] = {
632 #define PE_NO_ERROR 0
634 #define PE_UNSUPPORTED_SCHEME 1
635 N_("Unsupported scheme %s"), /* support for format token only here */
636 #define PE_MISSING_SCHEME 2
637 N_("Scheme missing"),
638 #define PE_INVALID_HOST_NAME 3
639 N_("Invalid host name"),
640 #define PE_BAD_PORT_NUMBER 4
641 N_("Bad port number"),
642 #define PE_INVALID_USER_NAME 5
643 N_("Invalid user name"),
644 #define PE_UNTERMINATED_IPV6_ADDRESS 6
645 N_("Unterminated IPv6 numeric address"),
646 #define PE_IPV6_NOT_SUPPORTED 7
647 N_("IPv6 addresses not supported"),
648 #define PE_INVALID_IPV6_ADDRESS 8
649 N_("Invalid IPv6 numeric address")
654 Return a new struct url if successful, NULL on error. In case of
655 error, and if ERROR is not NULL, also set *ERROR to the appropriate
658 url_parse (const char *url, int *error, struct iri *iri, bool percent_encode)
662 bool path_modified, host_modified;
664 enum url_scheme scheme;
667 const char *uname_b, *uname_e;
668 const char *host_b, *host_e;
669 const char *path_b, *path_e;
670 const char *params_b, *params_e;
671 const char *query_b, *query_e;
672 const char *fragment_b, *fragment_e;
675 char *user = NULL, *passwd = NULL;
677 const char *url_encoded = NULL;
678 char *new_url = NULL;
682 scheme = url_scheme (url);
683 if (scheme == SCHEME_INVALID)
685 if (url_has_scheme (url))
686 error_code = PE_UNSUPPORTED_SCHEME;
688 error_code = PE_MISSING_SCHEME;
692 if (iri && iri->utf8_encode)
694 iri->utf8_encode = remote_to_utf8 (iri, iri->orig_url ? iri->orig_url : url, (const char **) &new_url);
695 if (!iri->utf8_encode)
698 iri->orig_url = xstrdup (url);
701 /* XXX XXX Could that change introduce (security) bugs ??? XXX XXX*/
703 url_encoded = reencode_escapes (new_url ? new_url : url);
705 url_encoded = new_url ? new_url : url;
709 if (new_url && url_encoded != new_url)
712 p += strlen (supported_schemes[scheme].leading_string);
714 p = url_skip_credentials (p);
717 /* scheme://user:pass@host[:port]... */
720 /* We attempt to break down the URL into the components path,
721 params, query, and fragment. They are ordered like this:
723 scheme://host[:port][/path][;params][?query][#fragment] */
725 path_b = path_e = NULL;
726 params_b = params_e = NULL;
727 query_b = query_e = NULL;
728 fragment_b = fragment_e = NULL;
730 /* Initialize separators for optional parts of URL, depending on the
731 scheme. For example, FTP has params, and HTTP and HTTPS have
732 query string and fragment. */
733 seps = init_seps (scheme);
739 /* Handle IPv6 address inside square brackets. Ideally we'd
740 just look for the terminating ']', but rfc2732 mandates
741 rejecting invalid IPv6 addresses. */
743 /* The address begins after '['. */
745 host_e = strchr (host_b, ']');
749 error_code = PE_UNTERMINATED_IPV6_ADDRESS;
754 /* Check if the IPv6 address is valid. */
755 if (!is_valid_ipv6_address(host_b, host_e))
757 error_code = PE_INVALID_IPV6_ADDRESS;
761 /* Continue parsing after the closing ']'. */
764 error_code = PE_IPV6_NOT_SUPPORTED;
768 /* The closing bracket must be followed by a separator or by the
770 /* http://[::1]... */
772 if (!strchr (seps, *p))
774 /* Trailing garbage after []-delimited IPv6 address. */
775 error_code = PE_INVALID_HOST_NAME;
781 p = strpbrk_or_eos (p, seps);
784 ++seps; /* advance to '/' */
786 if (host_b == host_e)
788 error_code = PE_INVALID_HOST_NAME;
792 port = scheme_default_port (scheme);
795 const char *port_b, *port_e, *pp;
797 /* scheme://host:port/tralala */
801 p = strpbrk_or_eos (p, seps);
804 /* Allow empty port, as per rfc2396. */
805 if (port_b != port_e)
806 for (port = 0, pp = port_b; pp < port_e; pp++)
808 if (!c_isdigit (*pp))
810 /* http://host:12randomgarbage/blah */
812 error_code = PE_BAD_PORT_NUMBER;
815 port = 10 * port + (*pp - '0');
816 /* Check for too large port numbers here, before we have
817 a chance to overflow on bogus port values. */
820 error_code = PE_BAD_PORT_NUMBER;
825 /* Advance to the first separator *after* '/' (either ';' or '?',
826 depending on the scheme). */
829 /* Get the optional parts of URL, each part being delimited by
830 current location and the position of the next separator. */
831 #define GET_URL_PART(sepchar, var) do { \
833 var##_b = ++p, var##_e = p = strpbrk_or_eos (p, seps); \
837 GET_URL_PART ('/', path);
838 if (supported_schemes[scheme].flags & scm_has_params)
839 GET_URL_PART (';', params);
840 if (supported_schemes[scheme].flags & scm_has_query)
841 GET_URL_PART ('?', query);
842 if (supported_schemes[scheme].flags & scm_has_fragment)
843 GET_URL_PART ('#', fragment);
848 if (uname_b != uname_e)
850 /* http://user:pass@host */
852 /* uname_b uname_e */
853 if (!parse_credentials (uname_b, uname_e - 1, &user, &passwd))
855 error_code = PE_INVALID_USER_NAME;
860 u = xnew0 (struct url);
862 u->host = strdupdelim (host_b, host_e);
867 u->path = strdupdelim (path_b, path_e);
868 path_modified = path_simplify (scheme, u->path);
869 split_path (u->path, &u->dir, &u->file);
871 host_modified = lowercase_str (u->host);
873 /* Decode %HH sequences in host name. This is important not so much
874 to support %HH sequences in host names (which other browser
875 don't), but to support binary characters (which will have been
876 converted to %HH by reencode_escapes). */
877 if (strchr (u->host, '%'))
879 url_unescape (u->host);
880 host_modified = true;
882 /* Apply IDNA regardless of iri->utf8_encode status */
883 if (opt.enable_iri && iri)
885 char *new = idn_encode (iri, u->host);
890 host_modified = true;
896 u->params = strdupdelim (params_b, params_e);
898 u->query = strdupdelim (query_b, query_e);
900 u->fragment = strdupdelim (fragment_b, fragment_e);
902 if (opt.enable_iri || path_modified || u->fragment || host_modified || path_b == path_e)
904 /* If we suspect that a transformation has rendered what
905 url_string might return different from URL_ENCODED, rebuild
906 u->url using url_string. */
907 u->url = url_string (u, URL_AUTH_SHOW);
909 if (url_encoded != url)
910 xfree ((char *) url_encoded);
914 if (url_encoded == url)
915 u->url = xstrdup (url);
917 u->url = (char *) url_encoded;
923 /* Cleanup in case of error: */
924 if (url_encoded && url_encoded != url)
925 xfree ((char *) url_encoded);
927 /* Transmit the error code to the caller, if the caller wants to
934 /* Return the error message string from ERROR_CODE, which should have
935 been retrieved from url_parse. The error message is translated. */
938 url_error (const char *url, int error_code)
940 assert (error_code >= 0 && ((size_t) error_code) < countof (parse_errors));
942 if (error_code == PE_UNSUPPORTED_SCHEME)
945 char *scheme = xstrdup (url);
946 assert (url_has_scheme (url));
948 if ((p = strchr (scheme, ':')))
950 if (!strcasecmp (scheme, "https"))
951 error = aprintf (_("HTTPS support not compiled in"));
953 error = aprintf (_(parse_errors[error_code]), quote (scheme));
959 return xstrdup (_(parse_errors[error_code]));
962 /* Split PATH into DIR and FILE. PATH comes from the URL and is
963 expected to be URL-escaped.
965 The path is split into directory (the part up to the last slash)
966 and file (the part after the last slash), which are subsequently
970 "foo/bar/baz" "foo/bar" "baz"
971 "foo/bar/" "foo/bar" ""
973 "foo/bar/baz%2fqux" "foo/bar" "baz/qux" (!)
975 DIR and FILE are freshly allocated. */
978 split_path (const char *path, char **dir, char **file)
980 char *last_slash = strrchr (path, '/');
984 *file = xstrdup (path);
988 *dir = strdupdelim (path, last_slash);
989 *file = xstrdup (last_slash + 1);
992 url_unescape (*file);
995 /* Note: URL's "full path" is the path with the query string and
996 params appended. The "fragment" (#foo) is intentionally ignored,
997 but that might be changed. For example, if the original URL was
998 "http://host:port/foo/bar/baz;bullshit?querystring#uselessfragment",
999 the full path will be "/foo/bar/baz;bullshit?querystring". */
1001 /* Return the length of the full path, without the terminating
1005 full_path_length (const struct url *url)
1009 #define FROB(el) if (url->el) len += 1 + strlen (url->el)
1020 /* Write out the full path. */
1023 full_path_write (const struct url *url, char *where)
1025 #define FROB(el, chr) do { \
1026 char *f_el = url->el; \
1028 int l = strlen (f_el); \
1030 memcpy (where, f_el, l); \
1042 /* Public function for getting the "full path". E.g. if u->path is
1043 "foo/bar" and u->query is "param=value", full_path will be
1044 "/foo/bar?param=value". */
1047 url_full_path (const struct url *url)
1049 int length = full_path_length (url);
1050 char *full_path = xmalloc (length + 1);
1052 full_path_write (url, full_path);
1053 full_path[length] = '\0';
1058 /* Unescape CHR in an otherwise escaped STR. Used to selectively
1059 escaping of certain characters, such as "/" and ":". Returns a
1060 count of unescaped chars. */
1063 unescape_single_char (char *str, char chr)
1065 const char c1 = XNUM_TO_DIGIT (chr >> 4);
1066 const char c2 = XNUM_TO_DIGIT (chr & 0xf);
1067 char *h = str; /* hare */
1068 char *t = str; /* tortoise */
1069 for (; *h; h++, t++)
1071 if (h[0] == '%' && h[1] == c1 && h[2] == c2)
1082 /* Escape unsafe and reserved characters, except for the slash
1086 url_escape_dir (const char *dir)
1088 char *newdir = url_escape_1 (dir, urlchr_unsafe | urlchr_reserved, 1);
1092 unescape_single_char (newdir, '/');
1096 /* Sync u->path and u->url with u->dir and u->file. Called after
1097 u->file or u->dir have been changed, typically by the FTP code. */
1100 sync_path (struct url *u)
1102 char *newpath, *efile, *edir;
1106 /* u->dir and u->file are not escaped. URL-escape them before
1107 reassembling them into u->path. That way, if they contain
1108 separators like '?' or even if u->file contains slashes, the
1109 path will be correctly assembled. (u->file can contain slashes
1110 if the URL specifies it with %2f, or if an FTP server returns
1112 edir = url_escape_dir (u->dir);
1113 efile = url_escape_1 (u->file, urlchr_unsafe | urlchr_reserved, 1);
1116 newpath = xstrdup (efile);
1119 int dirlen = strlen (edir);
1120 int filelen = strlen (efile);
1122 /* Copy "DIR/FILE" to newpath. */
1123 char *p = newpath = xmalloc (dirlen + 1 + filelen + 1);
1124 memcpy (p, edir, dirlen);
1127 memcpy (p, efile, filelen);
1136 if (efile != u->file)
1139 /* Regenerate u->url as well. */
1141 u->url = url_string (u, URL_AUTH_SHOW);
1144 /* Mutators. Code in ftp.c insists on changing u->dir and u->file.
1145 This way we can sync u->path and u->url when they get changed. */
1148 url_set_dir (struct url *url, const char *newdir)
1151 url->dir = xstrdup (newdir);
1156 url_set_file (struct url *url, const char *newfile)
1159 url->file = xstrdup (newfile);
1164 url_free (struct url *url)
1170 xfree_null (url->params);
1171 xfree_null (url->query);
1172 xfree_null (url->fragment);
1173 xfree_null (url->user);
1174 xfree_null (url->passwd);
1182 /* Create all the necessary directories for PATH (a file). Calls
1183 make_directory internally. */
1185 mkalldirs (const char *path)
1192 p = path + strlen (path);
1193 for (; *p != '/' && p != path; p--)
1196 /* Don't create if it's just a file. */
1197 if ((p == path) && (*p != '/'))
1199 t = strdupdelim (path, p);
1201 /* Check whether the directory exists. */
1202 if ((stat (t, &st) == 0))
1204 if (S_ISDIR (st.st_mode))
1211 /* If the dir exists as a file name, remove it first. This
1212 is *only* for Wget to work with buggy old CERN http
1213 servers. Here is the scenario: When Wget tries to
1214 retrieve a directory without a slash, e.g.
1215 http://foo/bar (bar being a directory), CERN server will
1216 not redirect it too http://foo/bar/ -- it will generate a
1217 directory listing containing links to bar/file1,
1218 bar/file2, etc. Wget will lose because it saves this
1219 HTML listing to a file `bar', so it cannot create the
1220 directory. To work around this, if the file of the same
1221 name exists, we just remove it and create the directory
1223 DEBUGP (("Removing %s because of directory danger!\n", t));
1227 res = make_directory (t);
1229 logprintf (LOG_NOTQUIET, "%s: %s", t, strerror (errno));
1234 /* Functions for constructing the file name out of URL components. */
1236 /* A growable string structure, used by url_file_name and friends.
1237 This should perhaps be moved to utils.c.
1239 The idea is to have a convenient and efficient way to construct a
1240 string by having various functions append data to it. Instead of
1241 passing the obligatory BASEVAR, SIZEVAR and TAILPOS to all the
1242 functions in questions, we pass the pointer to this struct. */
1250 /* Ensure that the string can accept APPEND_COUNT more characters past
1251 the current TAIL position. If necessary, this will grow the string
1252 and update its allocated size. If the string is already large
1253 enough to take TAIL+APPEND_COUNT characters, this does nothing. */
1254 #define GROW(g, append_size) do { \
1255 struct growable *G_ = g; \
1256 DO_REALLOC (G_->base, G_->size, G_->tail + append_size, char); \
1259 /* Return the tail position of the string. */
1260 #define TAIL(r) ((r)->base + (r)->tail)
1262 /* Move the tail position by APPEND_COUNT characters. */
1263 #define TAIL_INCR(r, append_count) ((r)->tail += append_count)
1265 /* Append the string STR to DEST. NOTICE: the string in DEST is not
1269 append_string (const char *str, struct growable *dest)
1271 int l = strlen (str);
1273 memcpy (TAIL (dest), str, l);
1274 TAIL_INCR (dest, l);
1277 /* Append CH to DEST. For example, append_char (0, DEST)
1278 zero-terminates DEST. */
1281 append_char (char ch, struct growable *dest)
1285 TAIL_INCR (dest, 1);
1289 filechr_not_unix = 1, /* unusable on Unix, / and \0 */
1290 filechr_not_windows = 2, /* unusable on Windows, one of \|/<>?:*" */
1291 filechr_control = 4 /* a control character, e.g. 0-31 */
1294 #define FILE_CHAR_TEST(c, mask) (filechr_table[(unsigned char)(c)] & (mask))
1296 /* Shorthands for the table: */
1297 #define U filechr_not_unix
1298 #define W filechr_not_windows
1299 #define C filechr_control
1304 /* Table of characters unsafe under various conditions (see above).
1306 Arguably we could also claim `%' to be unsafe, since we use it as
1307 the escape character. If we ever want to be able to reliably
1308 translate file name back to URL, this would become important
1309 crucial. Right now, it's better to be minimal in escaping. */
1311 static const unsigned char filechr_table[256] =
1313 UWC, C, C, C, C, C, C, C, /* NUL SOH STX ETX EOT ENQ ACK BEL */
1314 C, C, C, C, C, C, C, C, /* BS HT LF VT FF CR SO SI */
1315 C, C, C, C, C, C, C, C, /* DLE DC1 DC2 DC3 DC4 NAK SYN ETB */
1316 C, C, C, C, C, C, C, C, /* CAN EM SUB ESC FS GS RS US */
1317 0, 0, W, 0, 0, 0, 0, 0, /* SP ! " # $ % & ' */
1318 0, 0, W, 0, 0, 0, 0, UW, /* ( ) * + , - . / */
1319 0, 0, 0, 0, 0, 0, 0, 0, /* 0 1 2 3 4 5 6 7 */
1320 0, 0, W, 0, W, 0, W, W, /* 8 9 : ; < = > ? */
1321 0, 0, 0, 0, 0, 0, 0, 0, /* @ A B C D E F G */
1322 0, 0, 0, 0, 0, 0, 0, 0, /* H I J K L M N O */
1323 0, 0, 0, 0, 0, 0, 0, 0, /* P Q R S T U V W */
1324 0, 0, 0, 0, W, 0, 0, 0, /* X Y Z [ \ ] ^ _ */
1325 0, 0, 0, 0, 0, 0, 0, 0, /* ` a b c d e f g */
1326 0, 0, 0, 0, 0, 0, 0, 0, /* h i j k l m n o */
1327 0, 0, 0, 0, 0, 0, 0, 0, /* p q r s t u v w */
1328 0, 0, 0, 0, W, 0, 0, C, /* x y z { | } ~ DEL */
1330 C, C, C, C, C, C, C, C, C, C, C, C, C, C, C, C, /* 128-143 */
1331 C, C, C, C, C, C, C, C, C, C, C, C, C, C, C, C, /* 144-159 */
1332 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
1333 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
1335 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
1336 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
1337 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
1338 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
1346 /* FN_PORT_SEP is the separator between host and port in file names
1347 for non-standard port numbers. On Unix this is normally ':', as in
1348 "www.xemacs.org:4001/index.html". Under Windows, we set it to +
1349 because Windows can't handle ':' in file names. */
1350 #define FN_PORT_SEP (opt.restrict_files_os != restrict_windows ? ':' : '+')
1352 /* FN_QUERY_SEP is the separator between the file name and the URL
1353 query, normally '?'. Since Windows cannot handle '?' as part of
1354 file name, we use '@' instead there. */
1355 #define FN_QUERY_SEP (opt.restrict_files_os != restrict_windows ? '?' : '@')
1357 /* Quote path element, characters in [b, e), as file name, and append
1358 the quoted string to DEST. Each character is quoted as per
1359 file_unsafe_char and the corresponding table.
1361 If ESCAPED is true, the path element is considered to be
1362 URL-escaped and will be unescaped prior to inspection. */
1365 append_uri_pathel (const char *b, const char *e, bool escaped,
1366 struct growable *dest)
1372 if (opt.restrict_files_os == restrict_unix)
1373 mask = filechr_not_unix;
1375 mask = filechr_not_windows;
1376 if (opt.restrict_files_ctrl)
1377 mask |= filechr_control;
1379 /* Copy [b, e) to PATHEL and URL-unescape it. */
1383 BOUNDED_TO_ALLOCA (b, e, unescaped);
1384 url_unescape (unescaped);
1386 e = unescaped + strlen (unescaped);
1389 /* Defang ".." when found as component of path. Remember that path
1390 comes from the URL and might contain malicious input. */
1391 if (e - b == 2 && b[0] == '.' && b[1] == '.')
1397 /* Walk the PATHEL string and check how many characters we'll need
1400 for (p = b; p < e; p++)
1401 if (FILE_CHAR_TEST (*p, mask))
1404 /* Calculate the length of the output string. e-b is the input
1405 string length. Each quoted char introduces two additional
1406 characters in the string, hence 2*quoted. */
1407 outlen = (e - b) + (2 * quoted);
1408 GROW (dest, outlen);
1412 /* If there's nothing to quote, we can simply append the string
1413 without processing it again. */
1414 memcpy (TAIL (dest), b, outlen);
1418 char *q = TAIL (dest);
1419 for (p = b; p < e; p++)
1421 if (!FILE_CHAR_TEST (*p, mask))
1425 unsigned char ch = *p;
1427 *q++ = XNUM_TO_DIGIT (ch >> 4);
1428 *q++ = XNUM_TO_DIGIT (ch & 0xf);
1431 assert (q - TAIL (dest) == outlen);
1434 /* Perform inline case transformation if required. */
1435 if (opt.restrict_files_case == restrict_lowercase
1436 || opt.restrict_files_case == restrict_uppercase)
1439 for (q = TAIL (dest); q < TAIL (dest) + outlen; ++q)
1441 if (opt.restrict_files_case == restrict_lowercase)
1442 *q = c_tolower (*q);
1444 *q = c_toupper (*q);
1448 TAIL_INCR (dest, outlen);
1451 /* Append to DEST the directory structure that corresponds the
1452 directory part of URL's path. For example, if the URL is
1453 http://server/dir1/dir2/file, this appends "/dir1/dir2".
1455 Each path element ("dir1" and "dir2" in the above example) is
1456 examined, url-unescaped, and re-escaped as file name element.
1458 Additionally, it cuts as many directories from the path as
1459 specified by opt.cut_dirs. For example, if opt.cut_dirs is 1, it
1460 will produce "bar" for the above example. For 2 or more, it will
1463 Each component of the path is quoted for use as file name. */
1466 append_dir_structure (const struct url *u, struct growable *dest)
1468 char *pathel, *next;
1469 int cut = opt.cut_dirs;
1471 /* Go through the path components, de-URL-quote them, and quote them
1472 (if necessary) as file names. */
1475 for (; (next = strchr (pathel, '/')) != NULL; pathel = next + 1)
1480 /* Ignore empty pathels. */
1484 append_char ('/', dest);
1485 append_uri_pathel (pathel, next, true, dest);
1489 /* Return a unique file name that matches the given URL as good as
1490 possible. Does not create directories on the file system. */
1493 url_file_name (const struct url *u)
1495 struct growable fnres; /* stands for "file name result" */
1497 const char *u_file, *u_query;
1498 char *fname, *unique;
1499 char *index_filename = "index.html"; /* The default index file is index.html */
1505 /* If an alternative index file was defined, change index_filename */
1506 if (opt.default_page)
1507 index_filename = opt.default_page;
1510 /* Start with the directory prefix, if specified. */
1512 append_string (opt.dir_prefix, &fnres);
1514 /* If "dirstruct" is turned on (typically the case with -r), add
1515 the host and port (unless those have been turned off) and
1516 directory structure. */
1519 if (opt.protocol_directories)
1522 append_char ('/', &fnres);
1523 append_string (supported_schemes[u->scheme].name, &fnres);
1525 if (opt.add_hostdir)
1528 append_char ('/', &fnres);
1529 if (0 != strcmp (u->host, ".."))
1530 append_string (u->host, &fnres);
1532 /* Host name can come from the network; malicious DNS may
1533 allow ".." to be resolved, causing us to write to
1534 "../<file>". Defang such host names. */
1535 append_string ("%2E%2E", &fnres);
1536 if (u->port != scheme_default_port (u->scheme))
1539 number_to_string (portstr, u->port);
1540 append_char (FN_PORT_SEP, &fnres);
1541 append_string (portstr, &fnres);
1545 append_dir_structure (u, &fnres);
1548 /* Add the file name. */
1550 append_char ('/', &fnres);
1551 u_file = *u->file ? u->file : index_filename;
1552 append_uri_pathel (u_file, u_file + strlen (u_file), false, &fnres);
1554 /* Append "?query" to the file name. */
1555 u_query = u->query && *u->query ? u->query : NULL;
1558 append_char (FN_QUERY_SEP, &fnres);
1559 append_uri_pathel (u_query, u_query + strlen (u_query), true, &fnres);
1562 /* Zero-terminate the file name. */
1563 append_char ('\0', &fnres);
1567 /* Check the cases in which the unique extensions are not used:
1568 1) Clobbering is turned off (-nc).
1569 2) Retrieval with regetting.
1570 3) Timestamping is used.
1571 4) Hierarchy is built.
1573 The exception is the case when file does exist and is a
1574 directory (see `mkalldirs' for explanation). */
1576 if ((opt.noclobber || opt.always_rest || opt.timestamping || opt.dirstruct)
1577 && !(file_exists_p (fname) && !file_non_directory_p (fname)))
1583 unique = unique_name (fname, true);
1584 if (unique != fname)
1588 /* On VMS, alter the name as required. */
1593 unique2 = ods_conform( unique);
1594 if (unique2 != unique)
1600 #endif /* def __VMS */
1605 /* Resolve "." and ".." elements of PATH by destructively modifying
1606 PATH and return true if PATH has been modified, false otherwise.
1608 The algorithm is in spirit similar to the one described in rfc1808,
1609 although implemented differently, in one pass. To recap, path
1610 elements containing only "." are removed, and ".." is taken to mean
1611 "back up one element". Single leading and trailing slashes are
1614 For example, "a/b/c/./../d/.." will yield "a/b/". More exhaustive
1615 test examples are provided below. If you change anything in this
1616 function, run test_path_simplify to make sure you haven't broken a
1620 path_simplify (enum url_scheme scheme, char *path)
1622 char *h = path; /* hare */
1623 char *t = path; /* tortoise */
1625 char *end = strchr (path, '\0');
1629 /* Hare should be at the beginning of a path element. */
1631 if (h[0] == '.' && (h[1] == '/' || h[1] == '\0'))
1636 else if (h[0] == '.' && h[1] == '.' && (h[2] == '/' || h[2] == '\0'))
1638 /* Handle "../" by retreating the tortoise by one path
1639 element -- but not past beggining. */
1642 /* Move backwards until T hits the beginning of the
1643 previous path element or the beginning of path. */
1644 for (--t; t > beg && t[-1] != '/'; t--)
1647 else if (scheme == SCHEME_FTP)
1649 /* If we're at the beginning, copy the "../" literally
1650 and move the beginning so a later ".." doesn't remove
1651 it. This violates RFC 3986; but we do it for FTP
1652 anyway because there is otherwise no way to get at a
1653 parent directory, when the FTP server drops us in a
1654 non-root directory (which is not uncommon). */
1663 /* A regular path element. If H hasn't advanced past T,
1664 simply skip to the next path element. Otherwise, copy
1665 the path element until the next slash. */
1668 /* Skip the path element, including the slash. */
1669 while (h < end && *h != '/')
1676 /* Copy the path element, including the final slash. */
1677 while (h < end && *h != '/')
1691 /* Return the length of URL's path. Path is considered to be
1692 terminated by one or more of the ?query or ;params or #fragment,
1693 depending on the scheme. */
1696 path_end (const char *url)
1698 enum url_scheme scheme = url_scheme (url);
1700 if (scheme == SCHEME_INVALID)
1701 scheme = SCHEME_HTTP; /* use http semantics for rel links */
1702 /* +2 to ignore the first two separators ':' and '/' */
1703 seps = init_seps (scheme) + 2;
1704 return strpbrk_or_eos (url, seps);
1707 /* Find the last occurrence of character C in the range [b, e), or
1708 NULL, if none are present. */
1709 #define find_last_char(b, e, c) memrchr ((b), (c), (e) - (b))
1711 /* Merge BASE with LINK and return the resulting URI.
1713 Either of the URIs may be absolute or relative, complete with the
1714 host name, or path only. This tries to reasonably handle all
1715 foreseeable cases. It only employs minimal URL parsing, without
1716 knowledge of the specifics of schemes.
1718 I briefly considered making this function call path_simplify after
1719 the merging process, as rfc1738 seems to suggest. This is a bad
1720 idea for several reasons: 1) it complexifies the code, and 2)
1721 url_parse has to simplify path anyway, so it's wasteful to boot. */
1724 uri_merge (const char *base, const char *link)
1730 if (url_has_scheme (link))
1731 return xstrdup (link);
1733 /* We may not examine BASE past END. */
1734 end = path_end (base);
1735 linklength = strlen (link);
1739 /* Empty LINK points back to BASE, query string and all. */
1740 return xstrdup (base);
1742 else if (*link == '?')
1744 /* LINK points to the same location, but changes the query
1745 string. Examples: */
1746 /* uri_merge("path", "?new") -> "path?new" */
1747 /* uri_merge("path?foo", "?new") -> "path?new" */
1748 /* uri_merge("path?foo#bar", "?new") -> "path?new" */
1749 /* uri_merge("path#foo", "?new") -> "path?new" */
1750 int baselength = end - base;
1751 merge = xmalloc (baselength + linklength + 1);
1752 memcpy (merge, base, baselength);
1753 memcpy (merge + baselength, link, linklength);
1754 merge[baselength + linklength] = '\0';
1756 else if (*link == '#')
1758 /* uri_merge("path", "#new") -> "path#new" */
1759 /* uri_merge("path#foo", "#new") -> "path#new" */
1760 /* uri_merge("path?foo", "#new") -> "path?foo#new" */
1761 /* uri_merge("path?foo#bar", "#new") -> "path?foo#new" */
1763 const char *end1 = strchr (base, '#');
1765 end1 = base + strlen (base);
1766 baselength = end1 - base;
1767 merge = xmalloc (baselength + linklength + 1);
1768 memcpy (merge, base, baselength);
1769 memcpy (merge + baselength, link, linklength);
1770 merge[baselength + linklength] = '\0';
1772 else if (*link == '/' && *(link + 1) == '/')
1774 /* LINK begins with "//" and so is a net path: we need to
1775 replace everything after (and including) the double slash
1778 /* uri_merge("foo", "//new/bar") -> "//new/bar" */
1779 /* uri_merge("//old/foo", "//new/bar") -> "//new/bar" */
1780 /* uri_merge("http://old/foo", "//new/bar") -> "http://new/bar" */
1784 const char *start_insert;
1786 /* Look for first slash. */
1787 slash = memchr (base, '/', end - base);
1788 /* If found slash and it is a double slash, then replace
1789 from this point, else default to replacing from the
1791 if (slash && *(slash + 1) == '/')
1792 start_insert = slash;
1794 start_insert = base;
1796 span = start_insert - base;
1797 merge = xmalloc (span + linklength + 1);
1799 memcpy (merge, base, span);
1800 memcpy (merge + span, link, linklength);
1801 merge[span + linklength] = '\0';
1803 else if (*link == '/')
1805 /* LINK is an absolute path: we need to replace everything
1806 after (and including) the FIRST slash with LINK.
1808 So, if BASE is "http://host/whatever/foo/bar", and LINK is
1809 "/qux/xyzzy", our result should be
1810 "http://host/qux/xyzzy". */
1813 const char *start_insert = NULL; /* for gcc to shut up. */
1814 const char *pos = base;
1815 bool seen_slash_slash = false;
1816 /* We're looking for the first slash, but want to ignore
1819 slash = memchr (pos, '/', end - pos);
1820 if (slash && !seen_slash_slash)
1821 if (*(slash + 1) == '/')
1824 seen_slash_slash = true;
1828 /* At this point, SLASH is the location of the first / after
1829 "//", or the first slash altogether. START_INSERT is the
1830 pointer to the location where LINK will be inserted. When
1831 examining the last two examples, keep in mind that LINK
1834 if (!slash && !seen_slash_slash)
1835 /* example: "foo" */
1837 start_insert = base;
1838 else if (!slash && seen_slash_slash)
1839 /* example: "http://foo" */
1842 else if (slash && !seen_slash_slash)
1843 /* example: "foo/bar" */
1845 start_insert = base;
1846 else if (slash && seen_slash_slash)
1847 /* example: "http://something/" */
1849 start_insert = slash;
1851 span = start_insert - base;
1852 merge = xmalloc (span + linklength + 1);
1854 memcpy (merge, base, span);
1855 memcpy (merge + span, link, linklength);
1856 merge[span + linklength] = '\0';
1860 /* LINK is a relative URL: we need to replace everything
1861 after last slash (possibly empty) with LINK.
1863 So, if BASE is "whatever/foo/bar", and LINK is "qux/xyzzy",
1864 our result should be "whatever/foo/qux/xyzzy". */
1865 bool need_explicit_slash = false;
1867 const char *start_insert;
1868 const char *last_slash = find_last_char (base, end, '/');
1871 /* No slash found at all. Replace what we have with LINK. */
1872 start_insert = base;
1874 else if (last_slash && last_slash >= base + 2
1875 && last_slash[-2] == ':' && last_slash[-1] == '/')
1877 /* example: http://host" */
1879 start_insert = end + 1;
1880 need_explicit_slash = true;
1884 /* example: "whatever/foo/bar" */
1886 start_insert = last_slash + 1;
1889 span = start_insert - base;
1890 merge = xmalloc (span + linklength + 1);
1892 memcpy (merge, base, span);
1893 if (need_explicit_slash)
1894 merge[span - 1] = '/';
1895 memcpy (merge + span, link, linklength);
1896 merge[span + linklength] = '\0';
1902 #define APPEND(p, s) do { \
1903 int len = strlen (s); \
1904 memcpy (p, s, len); \
1908 /* Use this instead of password when the actual password is supposed
1909 to be hidden. We intentionally use a generic string without giving
1910 away the number of characters in the password, like previous
1912 #define HIDDEN_PASSWORD "*password*"
1914 /* Recreate the URL string from the data in URL.
1916 If HIDE is true (as it is when we're calling this on a URL we plan
1917 to print, but not when calling it to canonicalize a URL for use
1918 within the program), password will be hidden. Unsafe characters in
1919 the URL will be quoted. */
1922 url_string (const struct url *url, enum url_auth_mode auth_mode)
1926 char *quoted_host, *quoted_user = NULL, *quoted_passwd = NULL;
1928 int scheme_port = supported_schemes[url->scheme].default_port;
1929 const char *scheme_str = supported_schemes[url->scheme].leading_string;
1930 int fplen = full_path_length (url);
1932 bool brackets_around_host;
1934 assert (scheme_str != NULL);
1936 /* Make sure the user name and password are quoted. */
1939 if (auth_mode != URL_AUTH_HIDE)
1941 quoted_user = url_escape_allow_passthrough (url->user);
1944 if (auth_mode == URL_AUTH_HIDE_PASSWD)
1945 quoted_passwd = HIDDEN_PASSWORD;
1947 quoted_passwd = url_escape_allow_passthrough (url->passwd);
1952 /* In the unlikely event that the host name contains non-printable
1953 characters, quote it for displaying to the user. */
1954 quoted_host = url_escape_allow_passthrough (url->host);
1956 /* Undo the quoting of colons that URL escaping performs. IPv6
1957 addresses may legally contain colons, and in that case must be
1958 placed in square brackets. */
1959 if (quoted_host != url->host)
1960 unescape_single_char (quoted_host, ':');
1961 brackets_around_host = strchr (quoted_host, ':') != NULL;
1963 size = (strlen (scheme_str)
1964 + strlen (quoted_host)
1965 + (brackets_around_host ? 2 : 0)
1968 if (url->port != scheme_port)
1969 size += 1 + numdigit (url->port);
1972 size += 1 + strlen (quoted_user);
1974 size += 1 + strlen (quoted_passwd);
1977 p = result = xmalloc (size);
1979 APPEND (p, scheme_str);
1982 APPEND (p, quoted_user);
1986 APPEND (p, quoted_passwd);
1991 if (brackets_around_host)
1993 APPEND (p, quoted_host);
1994 if (brackets_around_host)
1996 if (url->port != scheme_port)
1999 p = number_to_string (p, url->port);
2002 full_path_write (url, p);
2006 assert (p - result == size);
2008 if (quoted_user && quoted_user != url->user)
2009 xfree (quoted_user);
2010 if (quoted_passwd && auth_mode == URL_AUTH_SHOW
2011 && quoted_passwd != url->passwd)
2012 xfree (quoted_passwd);
2013 if (quoted_host != url->host)
2014 xfree (quoted_host);
2019 /* Return true if scheme a is similar to scheme b.
2021 Schemes are similar if they are equal. If SSL is supported, schemes
2022 are also similar if one is http (SCHEME_HTTP) and the other is https
2025 schemes_are_similar_p (enum url_scheme a, enum url_scheme b)
2030 if ((a == SCHEME_HTTP && b == SCHEME_HTTPS)
2031 || (a == SCHEME_HTTPS && b == SCHEME_HTTP))
2038 getchar_from_escaped_string (const char *str, char *c)
2040 const char *p = str;
2042 assert (str && *str);
2047 if (!c_isxdigit(p[1]) || !c_isxdigit(p[2]))
2055 return 0; /* error: invalid string */
2057 *c = X2DIGITS_TO_NUM (p[1], p[2]);
2058 if (URL_RESERVED_CHAR(*c))
2076 are_urls_equal (const char *u1, const char *u2)
2087 && (pp = getchar_from_escaped_string (p, &ch1))
2088 && (qq = getchar_from_escaped_string (q, &ch2))
2089 && (c_tolower(ch1) == c_tolower(ch2)))
2095 return (*p == 0 && *q == 0 ? true : false);
2099 /* Debugging and testing support for path_simplify. */
2102 /* Debug: run path_simplify on PATH and return the result in a new
2103 string. Useful for calling from the debugger. */
2107 char *copy = xstrdup (path);
2108 path_simplify (copy);
2114 run_test (char *test, char *expected_result, enum url_scheme scheme,
2115 bool expected_change)
2117 char *test_copy = xstrdup (test);
2118 bool modified = path_simplify (scheme, test_copy);
2120 if (0 != strcmp (test_copy, expected_result))
2122 printf ("Failed path_simplify(\"%s\"): expected \"%s\", got \"%s\".\n",
2123 test, expected_result, test_copy);
2126 if (modified != expected_change)
2128 if (expected_change)
2129 printf ("Expected modification with path_simplify(\"%s\").\n",
2132 printf ("Expected no modification with path_simplify(\"%s\").\n",
2136 mu_assert ("", modified == expected_change);
2141 test_path_simplify (void)
2144 char *test, *result;
2145 enum url_scheme scheme;
2148 { "", "", SCHEME_HTTP, false },
2149 { ".", "", SCHEME_HTTP, true },
2150 { "./", "", SCHEME_HTTP, true },
2151 { "..", "", SCHEME_HTTP, true },
2152 { "../", "", SCHEME_HTTP, true },
2153 { "..", "..", SCHEME_FTP, false },
2154 { "../", "../", SCHEME_FTP, false },
2155 { "foo", "foo", SCHEME_HTTP, false },
2156 { "foo/bar", "foo/bar", SCHEME_HTTP, false },
2157 { "foo///bar", "foo///bar", SCHEME_HTTP, false },
2158 { "foo/.", "foo/", SCHEME_HTTP, true },
2159 { "foo/./", "foo/", SCHEME_HTTP, true },
2160 { "foo./", "foo./", SCHEME_HTTP, false },
2161 { "foo/../bar", "bar", SCHEME_HTTP, true },
2162 { "foo/../bar/", "bar/", SCHEME_HTTP, true },
2163 { "foo/bar/..", "foo/", SCHEME_HTTP, true },
2164 { "foo/bar/../x", "foo/x", SCHEME_HTTP, true },
2165 { "foo/bar/../x/", "foo/x/", SCHEME_HTTP, true },
2166 { "foo/..", "", SCHEME_HTTP, true },
2167 { "foo/../..", "", SCHEME_HTTP, true },
2168 { "foo/../../..", "", SCHEME_HTTP, true },
2169 { "foo/../../bar/../../baz", "baz", SCHEME_HTTP, true },
2170 { "foo/../..", "..", SCHEME_FTP, true },
2171 { "foo/../../..", "../..", SCHEME_FTP, true },
2172 { "foo/../../bar/../../baz", "../../baz", SCHEME_FTP, true },
2173 { "a/b/../../c", "c", SCHEME_HTTP, true },
2174 { "./a/../b", "b", SCHEME_HTTP, true }
2178 for (i = 0; i < countof (tests); i++)
2180 const char *message;
2181 char *test = tests[i].test;
2182 char *expected_result = tests[i].result;
2183 enum url_scheme scheme = tests[i].scheme;
2184 bool expected_change = tests[i].should_modify;
2185 message = run_test (test, expected_result, scheme, expected_change);
2186 if (message) return message;
2192 test_append_uri_pathel()
2199 char *expected_result;
2201 { "http://www.yoyodyne.com/path/", "somepage.html", false, "http://www.yoyodyne.com/path/somepage.html" },
2204 for (i = 0; i < sizeof(test_array)/sizeof(test_array[0]); ++i)
2206 struct growable dest;
2207 const char *p = test_array[i].input;
2209 memset (&dest, 0, sizeof (dest));
2211 append_string (test_array[i].original_url, &dest);
2212 append_uri_pathel (p, p + strlen(p), test_array[i].escaped, &dest);
2213 append_char ('\0', &dest);
2215 mu_assert ("test_append_uri_pathel: wrong result",
2216 strcmp (dest.base, test_array[i].expected_result) == 0);
2223 test_are_urls_equal()
2229 bool expected_result;
2231 { "http://www.adomain.com/apath/", "http://www.adomain.com/apath/", true },
2232 { "http://www.adomain.com/apath/", "http://www.adomain.com/anotherpath/", false },
2233 { "http://www.adomain.com/apath/", "http://www.anotherdomain.com/path/", false },
2234 { "http://www.adomain.com/~path/", "http://www.adomain.com/%7epath/", true },
2235 { "http://www.adomain.com/longer-path/", "http://www.adomain.com/path/", false },
2236 { "http://www.adomain.com/path%2f", "http://www.adomain.com/path/", false },
2239 for (i = 0; i < sizeof(test_array)/sizeof(test_array[0]); ++i)
2241 mu_assert ("test_are_urls_equal: wrong result",
2242 are_urls_equal (test_array[i].url1, test_array[i].url2) == test_array[i].expected_result);
2248 #endif /* TESTING */